You are on page 1of 13

Science of the Total Environment 559 (2016) 326338

Contents lists available at ScienceDirect

Science of the Total Environment


journal homepage: www.elsevier.com/locate/scitotenv

Evaluating sustainable adaptation strategies for vulnerable mega-deltas


using system dynamics modelling: Rice agriculture in the Mekong Delta's
An Giang Province, Vietnam
Alexander Chapman , Stephen Darby
Geography and Environment, University of Southampton, Higheld, Southampton SO17 1BJ, UK

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

Mekong delta adaptation strategies


are evaluated with system dynamics
modelling.
For the rst time, sediment deposition
is integrated into the socioeconomic
system.
The shift towards high dykes and triplecropping benets land-wealthy farmers.
Mechanisms shown through which
triple-cropping forces debt on poorer
farmers.
Advantages of strategic facilitation of
sediment deposition are shown.

a r t i c l e

i n f o

Article history:
Received 8 January 2016
Received in revised form 22 February 2016
Accepted 23 February 2016
Available online 12 April 2016
Editor: D. Barcelo
Keywords:
Adaptation
Sediment
System dynamics
Rice
Mekong Delta

a b s t r a c t
Challenging dynamics are unfolding in social-ecological systems around the globe as society attempts to mitigate
and adapt to climate change while sustaining rapid local development. The IPCC's 5th assessment suggests these
changing systems are susceptible to unforeseen and dangerous emergent risks. An archetypal example is the
Vietnamese Mekong Delta (VMD) where the river dyke network has been heightened and extended over the
last decade with the dual objectives of (1) adapting the delta's 18 million inhabitants and their livelihoods to increasingly intense river-ooding, and (2) developing rice production through a shift from double to triplecropping. Negative impacts have been associated with this shift, particularly in relation to its exclusion of uvial
sediment deposition from the oodplain. A decit in our understanding of the dynamics of the rice-sediment system, which involve unintuitive delays, feedbacks, and tipping points, is addressed here, using a system dynamics
(SD) approach to inform sustainable adaptation strategies. Specically, we develop and test a new SD model
which simulates the dynamics between the farmers' economic system and their rice agriculture operations,
and uniquely, integrates the role of uvial sediment deposition within their dyke compartment.
We use the model to explore a range of alternative rice cultivation strategies. Our results suggest that the current
dominant strategy (triple-cropping) is only optimal for wealthier groups within society and over the short-term
(ca. 10 years post-implementation). The model suggests that the policy of opening sluice gates and leaving
paddies fallow during high-ood years, in order to encourage natural sediment deposition and the nutrient
replenishment it supplies, is both a more equitable and a more sustainable policy. But, even with this approach,
diminished supplies of sediment-bound nutrients and the consequent need to compensate with articial
fertilisers will mean that smaller-scale farmers in the VMD are more vulnerable to accruing debt.
2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).

Corresponding author.
E-mail address: adc506@gmail.com (A. Chapman).

http://dx.doi.org/10.1016/j.scitotenv.2016.02.162
0048-9697/ 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

A. Chapman, S. Darby / Science of the Total Environment 559 (2016) 326338

1. Introduction
Many of the world's deltas, which have a combined population of
over 500 million people (14% of the global total), are threatened with
major ooding and land loss as a consequence of rising relative sealevels (Syvitski et al., 2009). The displacement of people, the resultant
loss of lives and livelihoods, and declining food security are some of
the potential impacts of the drowning of the world's subsiding deltas
and these adverse impacts are signicant enough to have both regional
and global ramications (Kuenzer and Renaud, 2012; Warner, 2010).
Our understanding of the physical processes driving accelerated relative
sea-level rise (RSLR) in these vulnerable deltas is increasing. The net
rate of RSLR is now known to be controlled by the interplay between
processes that contribute to RSLR, namely: natural subsidence due to
the delta's compaction under its own weight; accelerated subsidence
due to anthropogenic inuences (e.g. groundwater and/or hydrocarbon
extraction); eustatic sea-level change; and those processes which can
slow or reverse RSLR, most notably aggradation as a result of the deposition of sediment supplied to the delta plain (Syvitski and Kettner,
2011). The unbalancing of these processes in deltas around the globe,
frequently in favour of processes that enhance RSLR, is driving exposure
vulnerability, a problem which is often exacerbated by the presence of
signicant local (social) sensitivity (Szabo et al., 2015; Tessler et al.,
2015).
With large areas of vulnerable deltas at risk, hard and soft adaptations are urgently required to mitigate the threat of RSLR. However, in
many deltas, particularly those located in developing nations, a policy
response is also required to tackle (and adapt to) other developmentlinked drivers of change, which include issues such as urbanisation,
agricultural intensication, and population growth (Kuenzer and
Renaud, 2012). Coping with the threats posed by these dual drivers of
change (environmental change and economic development), termed
double exposure (O'Brien and Leichenko, 2000), requires systemic understanding of the interacting social, economic, and physical components of the world's delta systems.
In highly complex social-ecological systems, such as river deltas, adaptations aiming to manage double exposure often involve trade-offs,
and some negative impacts may be unavoidable (Suckall et al., 2014).
Previous work has identied two particular risks in the adaptation process; actions may either have maladaptive traits themselves (Barnett
and O'Neill, 2010), or they may create undesirable impacts when they
interact with other development or climate change oriented actions,
creating a phenomenon termed emergent risk in the IPCC's 5th assessment report (Oppenheimer et al., 2014).
With the above risks in mind, adaptation to double exposure in
the world's deltas is receiving growing attention. For example,
Smajgl et al. (2015); Temmerman and Kirwan (2015), and Tessler
et al. (2015) have all recently emphasised the risks associated with
what they argue are the short term hard-engineering solutions
that are being pursued at present. For example, the use of river
dyke networks to protect agriculture and infrastructure from inundation has become common place (e.g. Hung et al., 2014; Hood,
2004; Nixon, 2003; Seto et al., 2002; Hensel et al., 1998; Ibez
et al., 1997). These dyke networks may provide valuable protection
from local ooding in the near term but, through their concurrent
exclusion of uvial sediment deposition (Manh et al., 2014) and
hence their adverse impact on RSLR, they can threaten the longterm sustainability of the delta-body (Syvitski et al., 2009). Specically, sediment deposition that is excluded by the presence of
dykes could otherwise mitigate exposure to ood and inundation
risks by offsetting land surface elevation reductions due to subsidence. However, an important but lesser discussed way in which
sediment deposition can additionally benet local social systems is
through the deposition of attached nutrients which subsequently
help maintain the productivity of ecosystems and agriculture (Olde
Venterink et al., 2006).

327

As a result of the potential negative impacts associated with hardengineering interventions, alternative nature-based solutions that involve the strategic facilitation of oodplain sediment deposition to
build land height and counter RSLR are increasingly being advocated
as an alternative management approach. This concept of facilitated sediment deposition, which emerged in the more developed Mississippi
and Ebro Deltas (Rovira and Ibez, 2007; Day et al., 2005), has now
been recognised more widely as a potential adaptation strategy
(Ibez et al., 2014) and is being incorporated into the management
plans of developing deltas including large, highly vulnerable, deltas
such as the Mekong, through the Mekong Delta Plan (MDP, 2013).
However, such nature-based adaptation strategies have not yet been
systemically evaluated against the hard-engineering, dyke-based, alternatives. For example, from the perspective of some stakeholders, such
nature-based strategies may conict with other local non-adaptation
oriented objectives, such as achieving maximum short-term agricultural production and/or ood protection.
While our understanding of the physical processes involved in oodplain sediment dynamics and their interaction with hydraulic infrastructure has been developing in recent years (e.g. Manh et al., 2014;
Hung et al., 2014), complementary knowledge of the socioeconomic
trade-offs linking them to agriculture remains lacking. This knowledge
decit broadly includes the value and role of the free fertilising nutrients
bound to the uvial sediments deposited in deltas, and the optimal socioeconomic balance between ood exclusion and facilitation. In many
deltas, decisions on oodplain sediment management are intrinsically
linked to agricultural development objectives and management decisions that affect the livelihoods of many millions of farmers. In the
Asian mega-deltas in particular, comparatively poor farmers have become dependent for their subsistence on the stable environment provided by hydraulic infrastructure, and the high productivity of soils,
creating a phenomenon that Evers and Benedikter (2009) have termed
a modern hydraulic society. Recent research has highlighted the important role that wealth inequalities can play in determining how well
deltaic societies cope with environmental change (Szabo et al., 2015).
As a result, adaptation planning must be conducted with systematic
consideration of the potential impacts of different implementation
strategies and paying particular attention to the factors which determine a strategy's winners and losers.
To address some of these issues, in this paper we examine socioeconomic dynamics linking local livelihoods, hydraulic infrastructure, management practices, uvial inundation, and sediment deposition in the
Mekong delta. The Mekong delta (introduced further in Section 3) is
currently facing a major sustainability challenge (Anthony et al.,
2015). As the world's third largest delta (Coleman and Huh, 2003) it
hosts a population of nearly 20 million, but its signicance for the present study lies primarily in its pivotal importance to the food security of
the Southeast Asia region more generally. Specically, the delta provides 50% of Vietnam's food (GSO, 2014) and over 90% of Vietnam's
rice production (Kontgis et al., 2015). Not only does rice production
form the foundation of the local economy (as discussed in Section 3)
but, as Mohanty et al. (2013) outline, rice is the staple food product of
hundreds of millions of the world's poorest people. Globally, there are
signicant threats to rice production as it is primarily concentrated in
a small number of low-lying and/or ood-prone areas (e.g. deltas), making rice production particularly vulnerable to the impacts of environmental change (Wassmann et al., 2009). Our focus on the Mekong
may therefore have relevance to the world's other major rice producing
deltas.
To investigate the trade-offs involved in adaptation and agricultural
development management decisions we evaluate the competing merits
of different policy scenarios by developing, testing and applying a novel
system dynamics model (SDM). The new SDM is, to our knowledge, the
rst that is capable of integrating the role of deltaic sediment-bound nutrient deposition into the economic and decision making systems of Mekong delta rice farmers. We employ the SDM to evaluate the system

328

A. Chapman, S. Darby / Science of the Total Environment 559 (2016) 326338

dynamics, such as thresholds, feedbacks, and delays, of the recent


transition (adaptation) from the prior low dyke network to
sediment-excluding high dykes, and especially the shift from the
double to triple rice-cropping production system which has resulted
(Kontgis et al., 2015). As the rst such study, our work has relevance
to a range of delta environments where there are urgent needs for
evaluations of different adaptation policies, but particularly contexts
where sediment exclusion due to investment in hydraulic infrastructure may be affecting rice cultivation. Within the specic context of
the Mekong delta, our work has policy relevance as our lack of understanding of the socioeconomic role of sediment in the double to
triple-cropping switch has been identied in the Mekong Delta
Plan (2013, p.24), a key national policy analysis document, as a specic knowledge gap.
2. Aims and objectives
In this paper we aim to perform a comparative evaluation of a
suite of adaptation strategies for the agricultural sector of the
Vietnamese Mekong delta (as represented by the exemplar of rice
cultivation in An Giang province). Specically, we apply a system
dynamics methodology to explore the dynamics of different courses
of action implemented in a context of double exposure to environmental change and development pressures. The following specic
objectives are addressed:
- To identify the key changes in socioeconomic system dynamics
resulting from the shift (adaptation) from low to high river dykes,
and the associated exclusion of uvial sediment to rice paddies.
- To determine the effectiveness of alternative rice production policies
which have been implemented and proposed.
- To evaluate how the different policies analysed operate and perform
across farmers within different wealth strata.

3. Study area
The VMD (Fig. 1A) is a highly productive agricultural region that has
been undergoing rapid economic development since the Doi Moi opening of the Vietnamese markets in 1986. In large part the VMD's development has been driven by increases in rice production and export (ISGMARD, 2011). Expansion and intensication have been facilitated by
improvements in farming practice, uptake of modern rice varieties, increased input levels, and multiplication of crops (Garschagen et al.,
2012). However, the technological development of agriculture in the region remains relatively low and the largest single cost in the rice production process remains articial fertilisation, constituting around 40
50% of overall expenditure (Pham, 2010, p.230). Despite the region's
rapid development, poverty is still prevalent and farmers face multiple
challenges, such as declining productivity, income insecurity, and debt
(Garschagen et al., 2012; Swain et al., 2008). As of 2012, the General Statistics Ofce of Vietnam (GSO, 2014) estimated that the poverty rate in
the VMD was 16.2%. Poverty in this context was dened by an individual
income threshold (at 2012 prices) of less than approximately US$500
per year (Demombynes and Linh, 2015).
As discussed above, the Mekong delta is one of the world's largest
rice producing regions, with much of the crop being exported. As such
the Mekong delta is strategically important in terms of both the
Vietnamese economy and global food security (Smajgl et al., 2015). An
Giang province (Fig. 1B), located in the northern part of the
Vietnamese delta, has been selected as the focus of this study as it can
be considered typical of those parts of the delta where rice production
is the dominant economic activity, with 85% of land use devoted to
rice (AGSO, 2013). The intensication of rice production in the delta
(Kontgis et al., 2015) has required ever increasing control of the uvial
oods which would otherwise inundate the majority of the northern
part of the delta. Typically, rice production takes place within rectangular dyke rings (Fig. 1B) which protect multiple smallholder rice paddies

Fig. 1. A.) The provinces and agro-ecological zones of the Vietnamese Mekong Delta. Highlighted are the names of the nine surveyed communes and inset is the Mekong river basin which
supplies the sediment entering the delta from the north. B.) An Giang province and the high dyke network which represents the adaptation action undertaken. The boundaries of the nine
surveyed communes are also indicated. Data on the location and extent of the dyke network as of 2007 was provided by the WISDOM project (2014). The WISDOM project dyke height
estimates were updated using agricultural data contained within the An Giang Statistical Yearbook (AGSO, 2013) to estimate their 2013 extent.

A. Chapman, S. Darby / Science of the Total Environment 559 (2016) 326338

from the annual monsoonal ood and supply irrigation water yearround. However, these ring dyke networks also dramatically affect the
inundation regime. Fluvial oods are now subject to a second key driver
of change, climate change. Of the multiple climate-change risks present
in the VMD, increases in the frequency and intensity of uvial and coastal ooding stand-out as particular threats to the wellbeing of the delta's
inhabitants and those dependent on its economy (Van et al., 2012). The
structures (dykes and sluice gates) previously built to control uvial
ooding for the safety of residents and crops have now been adopted
as important components of the government's strategy for adapting to
climate change (Vietnamese Government, 2011; MARD, 2008; MPI,
2006). An Giang province is representative of much of the rest of the
Vietnamese Mekong delta in having undergone a large programme of
dyke height increases from low (02 m) to high (3.5 m +) since the
new millennium.
However, a growing body of evidence associates ring dyke networks
with both positive and negative impacts on local communities and the
wider delta (Birkmann et al., 2012; Garschagen et al., 2012; Pham,
2011; Sakamoto et al., 2009). On the one hand, high dykes help meet
both development and adaptation objectives by facilitating triplecropping and protecting livelihoods against intensifying uvial ooding
and inundation driven by sea-level rise; on the other, the exclusion of
sediment (and associated nutrients) reduces agricultural productivity
in the long term, and accelerates the rate of RSLR. The result is a nexus
of conicting long and short-term development and adaptation objectives and pressures which create an archetypal example of a system
with potential for emergent risk.
3.1. Rice cultivation systems in the Mekong delta
At present, two primary farming systems operate within An Giang's
dyke rings, double and triple rice-cropping, which are split approximately 1:2 by area (Kontgis et al., 2015), inclusive of one variant on
triple-cropping, the 3-3-2 cycle (introduced below). The doublecropping pattern is almost always found within low (02 m) dyke
rings and was the dominant practice during the 1990s. In such dyke
rings ooding can still take place when peak water heights overow
the dykes during the monsoon season. During these periods of ooding
suspended sediment may be deposited on the oodplain (Hung et al.,
2014); the quality and quantity of the deposited sediments varying spatially and inter-annually (Manh et al., 2014). Aware that the sediment
deposits have many of the nutrients required for rice agriculture (estimated at 50% of the total N, P, and K required for a crop; Manh et al.,
2014) double-cropping farmers have adopted the practice of spreading
the deposited sediment evenly over their paddy. Moreover, farmers also
factor the quantity of sediment they perceive as having been deposited
into a wider decision making process regarding the quantity of articial
fertiliser they apply in a given year. This decision making process is
based on a combination of local and historical knowledge, crude tools
for measuring crop-health, and the farmer's allocative efciency (i.e.
the farmer's assessment of the optimal choice based on current prices
and yields). Farmers may apply a certain quantity of fertiliser, but the
total quantity of nutrients reaching the plant is then subject to their
technical efciency and the technology at their disposal (Khai and
Yabe, 2011). Once a rice crop has been harvested the majority is sold
for export, though a small quantity may be kept back for personal consumption. Traditionally, farmers employing the double-cropping system will also seek income from secondary sources, such as shing or
small-scale livestock husbandry. This is because, under doublecropping, farmers have approximately a four month period (JulOct)
during the monsoon season when their paddy is left fallow and they
can focus on non-rice income generation.
Within the area enclosed by high (2 m+) dyke rings, an area which
almost doubled between 2000 and 2010 (Kontgis et al., 2015), ooding
either does not take place at all or comes via sluice gate operation,
reducing the natural (i.e. pre-high dyke) rate of sediment deposition

329

(Hung et al., 2014). The exclusion of the annual ood also reduces
local shing potential and, perceiving a fallow compartment and subject
to government production targets, farmers are incentivised to grow a
third rice crop potentially increasing annual yield per hectare by a
factor of 1.5. The additional crop results in a considerably increased
workload and hence reduces the triple-cropping farmer's capacity to
generate income from other sources (Garschagen et al., 2012). Furthermore, the loss of the ushing effect of the monsoonal ood is cited as a
factor contributing to proportionally higher pest and disease costs
(Pham, 2011). Aware of some of these negative impacts, the An Giang
provincial government recommend adoption of a variant 3-3-2
cropping cycle, in which the paddy is left fallow and a full ooding season is facilitated by sluice gate operation every third year. However, this
recommended strategy has thus far seen low uptake (Sakamoto et al.,
2009) and was found in only 2 of 9 An Giang communes in a 2014 survey conducted by Chapman et al. (2016).
4. Model construction
In environmental systems, such as the VMD, which are subject to
high levels of forecasting uncertainty and which are heavily inuenced by local management decisions a predict-then-act (Lempert
et al., 2004) approach to adaptation decision making can be
problematic. Uncertainty and complexity can mean that identifying
a single preferential adaptation strategy can be challenging and
furthermore, can enhance the desirability of short-term, palliative,
solutions (Fssel, 2007). As a result, interest has grown in robust
decision making (Lempert et al., 2006). For robust decision making
knowledge is required about the operational sources of vulnerability
within a social-ecological system, at which action can be targeted
(Costa et al., 2011; Fraser et al., 2011). System dynamics modelling
(SDM) enables complex interactions between system components
to be evaluated through exploration of how stocks (i.e. accumulation
points, such as the stock of nutrients held within the oodplain) and
ows (also called rates, as represented by the differential equations
which dene how the level of a stock varies over time) connect to
form a system. SDMs are not usually regarded as forecasting tools,
but rather as tools that enable users to investigate system dynamics,
identify unforeseen outcomes and compare scenarios of action or
system change (Ford, 2010, p. 5960; Simonovic and Li, 2004).
These three strengths have begun to see the SDM methodology
being applied to the task of climate change mitigation and adaptation policy evaluation (e.g. Dace et al., 2015; Gies et al., 2014) and,
more generally, to evaluating sustainable policy in coupled hydrological and economic systems (e.g. Alcal et al., 2015; Sunik et al.,
2013).
In the following sub-sections we outline how our SDM was constructed and tested for the Mekong delta, focusing on the
(i) identication of system processes, (ii) representation of stocks and
ows, (iii) evaluation of the model, and (iv) the formulation of the specic policy scenarios evaluated herein.
4.1. Identication of system processes
Scoping eld studies were performed in 2013 and 2014 during
which semi-structured interviews were conducted with six local academics, four senior provincial policy makers, and eighteen commune
authority leaders to identify expert and stakeholder views on key aspects of system behaviour within the VMD rice producing regions. The
knowledge gained was used to build a causal loop diagram (CLD) of
the system in question (Fig. 2). Fig. 2 highlights how interactions between three key sectors of the system (physical, economic, and decision
making) are analysed. Fig. 2 also presents the core feedback loops which
were identied through the process of stakeholder engagement and
included in the model, as well as one example of a feedback loop that
was regarded as lying outside the boundaries of the system under

330

A. Chapman, S. Darby / Science of the Total Environment 559 (2016) 326338

Fig. 2. A casual loop diagram (CLD) showing modelled connections, their nature, individual polarity (positive or negative), and the overall loop polarity (sometimes termed the momentum
of the loop, which can either be reinforcing + or balancing ). Included is one example of a feedback loop excluded from the model, with this exception, this diagram represents the
boundaries of the model.

examination and therefore excluded from the model. Key exogenous


forcing variables that were included in the model were: increases and
unpredictability in fertiliser and rice prices, increases in rice yield due
to new variety development, unpredictability in rice yield due to weather variability, change and variability in the quantity of sediment-bound
nutrients deposited annually by uvial inundation, and increases and
unpredictability in the costs associated with pests and disease (particularly those associated with the triple-cropping production regime).

4.2. Stock and ow design


The model's stock and ow structure was designed and parameters
calibrated utilising a combination of the knowledge gained in keyinformant interviews and through a comprehensive literature search.
Each of the forcing variables listed above were assigned a rate of change
and variability (standard deviation) the values of which were extracted
from secondary data sources (listed and described in Table 1) using

Table 1
The (exogenous) model driving data and the values for each which were used to initiate the model, their sources, information type, and an assessment of the source's reliability. Overall
score represents the aggregated source reliability score based on the system set out in the Supplementary Materials. Highlighted, are the components which were taken forward for sensitivity analysis.

Data

Source (s)

Information Type

Units

Initial
value

Overall
score

Interannual variability of suspended


sediment concentration

SIWRR, 2013 (daily time series 20052011);


Shreshtha et al. 2013 (predictions)

Statistical/
Modelled

Fraction (standard
deviation)

0.2

89%

Dam trapping efficiencies (end of


simulation sediment reduction)

Kummu et al. 2010; Kondolf et al. 2014

Modelled

Fraction

0.84

69%

Total nutrient contentof suspended


sediment (N, P, K)

Manh et al. 2014

Uncontrolled

Kg/ha/yr

300

78%

Variability of rice prices

FAO, 2014 (monthly time series 20082014),


Survey data

Statistical

Fraction (standard
deviation)

0.05

78%

Rate of change of rice prices

FAO, 2014 (monthly time series 20082014)

Statistical

%/ha/yr/yr

78%

Variability of fertiliser prices

World Bank, 2014 (monthly time series 2000


2014), Survey data

Statistical

Fraction (standard
deviation)

0.1

78%

Rate of change of fertiliser prices

World Bank, 2014 (monthly time series 2000


2014)

Statistical

%/ha/yr/yr

1.05

33%

Growth rate of rice yields due to rice


variety development

Laborte et al. 2012; Ray et al. 2012; Tran and


Kajisa, 2006

Statistical/Case
study

%/ha/yr/yr

1.01

89%

Nonrice income achievable

Bosma et al. 2005

Statistical

000 VND/yr

4800

78%

0.1

78%
78%

Exogenous variability of rice yield

GSO, 2014 (seasonal time series 19952013)

Statistical

Fraction (standard
deviation)

Rice price

Survey data

Statistical

000 VND/Kg

4.5

Fertiliser price

Survey data

Statistical

000 VND/Kg

78%

Minimum wage level

Vietnamese Government, 2013

Law

000 VND/person/yr

7200

100%

A. Chapman, S. Darby / Science of the Total Environment 559 (2016) 326338

331

Table 2
Endogenous modelled processes (key micro-systems within the model), their sources, information type, and an assessment of their source's reliability. Overall score represents the aggregated source reliability score based on the system set out in the supplementary materials. Highlighted, are the components which were taken forward for sensitivity analysis.

Modelled processes

Source(s)

Information Type

Overall score

Sediment deposition process

Manh et al. 2014; Hung et al. 2014

Uncontrolled

75%

Floodplain nutrient accumulation process

Tsheboeng et al., 2014

Uncontrolled

58%
67%

Nutrient fixing and leaching process

Liang et al., 2013; Phong et al., 2011

Controlled/Uncontrolled

Rice nutrient requirement (production function)

Pham et al. 2004; Witt et al. 1999

Controlled

83%

Technological advancement process

Reardon et al. 2014; Rutten et al., 2014

Statistical/Modelled

58%

Technical efficiency rate

Khai and Yabe, 2011; Hueglas and Templeton,


2010

Statistical

75%

Technological investment return

Tin et al., 2008; Tran et al., 2000

Statistical

75%

Sediment perceived versus fertiliser applied decision


process

Survey data

Stakeholder

100%
83%

Fixed cost variation between cropping patterns

Survey data

Stakeholder

Pesticide cost variation between cropping patterns

Survey data

Stakeholder

83%

Fertiliser subsidy policy

Tran, 2014

Expert

67%

Farmers propensity to invest

Personal Intuition

Personal Intuition

Farmers fraction of funds kept as contingency

Personal Intuition

Personal Intuition

simple statistical analysis; their initial values are summarised in Table 1.


As is common in SDM projects a variety of different data sources were
used to inform the design of the stock and ow structures used to create
the model; the processes modelled and their sources are outlined in
Table 2. Our model was developed using ISEE Systems' IThink SDM software package.
4.3. Evaluation of model performance
As noted above, SDMs are built with the objective of understanding a
system to address a problem or answer a question, and not to predict or
forecast future conditions (Sterman, 2000). System dynamics models
should therefore be evaluated with this purpose in mind. As such, in
the following sub-sections we describe three key phases of our model
evaluation process, which follows Sterman's (2000) Tests for assessment of dynamic models.
4.3.1. Parameter assessment
In the rst phase of the model evaluation process, termed the
judgemental parameter assessment, all of the sources informing relationships in the model were assigned a score for their strength in
terms of: statistical condence (where applicable), study location transferability, spatial scale comparability, and quantity of evidence/studies
(see Supplementary Information for further details on this evaluation
process; Tables 1 and 2 present the average reliability scores of the exogenous data sources and the modelled relationships, respectively).
The process of evaluating condence in data sources identied ve
key parameters against which there was notably greater uncertainty
(scoring one standard deviation below the mean or more): (i) the
fertiliser price rate of change (%/season), (ii) the farmers' propensity
to invest for the future (% prot/season), (iii) the fraction of the farmers'
incomes kept as contingency (% prot/season), (iv) the time from
sediment-bound nutrient deposition to availability for plant uptake
(seasons), and (v) the rate of depreciation of farming technology investments (%/season). These ve parameters were therefore considered
further during the model evaluation phase which addressed the model's
sensitivity, as described in Section 4.3.3.

(Chapman et al., 2016; Fig. 1 shows the locations of the communes


within which the survey took place). It should be recognised that
these observational data are subject to their own uncertainty, particularly as the survey required farmers to recall information spanning a
six year historical period (20082013). We compared simulated and
observed time-series of the yield/fertiliser ratio achieved by farmers operating three different farming systems: (i) double-cropping, (ii) triplecropping, and (iii) those farmers who changed cropping system during
the observation period. The yield/fertiliser ratio was selected as the primary metric for model comparison because its temporal trends succinctly summarise the status and sustainability of an agricultural
system and it is commonly utilised to benchmark the performance of
policies implemented in agricultural systems (see Khai and Yabe, 2011).
When simulating the yield/fertiliser ratio time-series under the
three key system conditions listed above (i-iii) the model produced relative errors (RE) against the reference mode that varied between 0.1
and 7.3% (Table 3). There is evidence (p b 0.05) of a systematic error
in all of the comparisons of the absolute values of the yield/fertiliser
ratio achieved by farmers of around 1 t/t (5.37.3%). When comparing
rates of change there is no statistically signicant (p b 0.05) evidence
of a systematic error, the model predicts the correct signal in all three
cases, and RE ranges from 0.13.9%. The model would appear to provide
robust simulations of the two crop system but, overestimates the
reported values under the triple-cropping system.
4.3.3. Sensitivity analysis
Sensitivity analyses were conducted on the ve parameters identied in Section 4.3.1 as having notably weak evidence bases. In doing
so our objective was to investigate the potential for such weaknesses
to confound our overall condence in the model simulations. We used

Table 3
A summary of the results of the statistical tests used to validate the model outputs against
the farmer-reported data. SE = systematic error (reported data modelled data), RE =
relative error.
Cropping pattern

Absolute Yielfert values


SE (t/t)

4.3.2. Comparison with reference mode


The second phase of evaluation involved undertaking statistical
comparisons of SDM derived data versus real-world data (the latter
being termed the reference mode by Ford, 2010) obtained from a comprehensive 2014 survey of 195 An Giang rice farming households

Two-crop
During cropping
system change
Three-crop

Rate of change of Yielfert

SE
RE SE (t/t/yr)
p-value (%)

SE
RE
p-value (%)

0.95 0.58 b0.001


1.14 1.00
0.013

5.3
6.3

0.001 0.023 0.999


0.039 0.039 0.058

0.1
3.9

1.36 0.70 b0.001

7.3

0.019 0.028 0.354

1.9

332

A. Chapman, S. Darby / Science of the Total Environment 559 (2016) 326338

Table 4
A summary of the general sensitivity degrees, GSQ, between ve parameters and four output variables: desired level of fertilisation (DF); rice yield (RY); cash prot (CP); and technical efciency (TE). Values N0.1 are highlighted in bold.
Indicator
DF

RY

CP

TE

Fertiliser price rate of change


Farmer's propensity to invest
Farmer's fraction of funds kept as contingency
Time to nutrient availability
Technological depreciation rate

0.085
0.035
0.008
0.015
0.051

0.011
0.003
0.002
0.000
0.008

0.65
0.12
0.001
0.017
0.004

0.048
0.034
0.006
0.003
0.044

the sensitivity degree (SQ), a metric commonly applied in SDM contexts


and shown in Eq. (1), to measure the sensitivity of four key indicators
against incrementally increased levels of variation (10%) in the ve
parameters tested. The general sensitivity degree (GSQ) represents the
average of all sensitivity degrees (SQ) measured for the different levels
of perturbation of a particular parameter.



Q
t X t
SQ
:

Q t X t

(Zhang et al., 2008)Q(t) = Model output at time t; X(t) = Model input at


time t; SQ = Sensitivity degree.
The majority of the tested combinations revealed low (b 0.1) general
sensitivity (Table 4). However, a notable exception is that increases in
the fertiliser price rate of change were seen to have a considerable inuence over the farmer's cash prot (CP). The time-dependence of the
sensitivity of this key relationship is explored in Table 5. The sensitivity
of cash prot to the fertiliser price rate of change grows throughout the
test simulation such that, by the end of the simulation, its inuence on
CP was proportionately greater than the initial change made to the
parameter. Considering the unpredictable nature of the fertiliser market
it is unlikely that any greater condence in this particular parameter is
attainable. However, rather than invalidating the model, this nding
might be regarded as offering insight into the nature of the system; as
such it is investigated further in Section 5.1.
4.4. Scenario design and testing
All simulations conducted herein were initialised with those forcing
parameters which change throughout the simulation (e.g. fertiliser
price) set below the levels at which policy testing began (equivalent
to the 2013 averages see Supplementary Information) and a spinup simulation period was used to bring the system up to the testing initiation point, i.e. average 2013 levels. The model took a spin-up period of
53 time steps (seasons of which there are three seasons per calendar
year) to reach this policy initiation point. The spin-up was conducted
with the model parameters set to simulate the double-cropping rice
production system which has historically operated within the study region. At the end of the spin-up period minor adjustments were made to
the model's parameters to congure the model for a specic ricecultivation policy (as detailed in Table 6) and the model was then run

Table 5
A breakdown of the sensitivity degrees (SQ) of the parameter representing fertiliser price
increase across various time steps and levels of alteration.

10%
20%
30%
40%

Policy Description and background


1

Parameter

Change in parameter level

Table 6
A summary of the policy scenarios tested with the model. Each policy was implemented
through manipulation of the parameter cropping pattern.

Time step
53

73

93

113

0.064
0.062
0.060
0.059

0.42
0.47
0.84
0.67

0.92
0.91
0.72
0.67

1.7
1.4
0.99
0.91

2
3
4

Business-as-usual for triple-cropping farmers (two-thirds of AG farmers);


at the end of the spin-up farmers switch to and operate the triple-cropping
system
Business-as-usual for (one-third of AG) farmers who remain operating the
old double-cropping system
Sluice gates are opened to allow oodplain sediment deposition once
every three years (the 3-3-2 system)
Sluice gates are opened to allow sediment deposition in years of high ood
and sediment deposition potential (a strategy advocated by farmers
interviewed). Sediment deposition rates were smoothed over 12 model
time steps (4 years) and deposition was allowed only in years exceeding
10% above the smooth.

over a period of 20 further years (i.e. 60 model time steps/seasons),


the testing period. With the number of forcing variables affecting the
system and our stated aim not to forecast the system's conditions, it
was felt that moving any further into the future could potentially render
our outputs meaningless, while 60 time steps is sufciently long for the
system dynamics to emerge. The model constructed represents the
VMD system as we presently understand it and with a wide variety of
parameters making up the model it was important to reduce the risk
of issues such as overtting affecting the predicted data.
Four policy scenarios are analysed herein (Table 6, see also
Section 3.1). Three of those scenarios simulated rice cropping systems
currently operated in the region: (1) the triple-cropping system (business-as-usual); (2) the double-cropping system (the traditional system); and (3) the 33-2 cropping rotation that has been implemented
in some districts to alleviate a negative impact (sediment exclusion)
of the triple-cropping system. Additionally, one further policy not currently operated in the region was tested. This policy (4) is consistent
with the recommendations made by the MDP (2013) in which triplecropping continues but with strategic ooding (and double-cropping)
during years of high ood and sediment deposition potential. Notably,
policy 4 was also informally described as being preferred by many of
the farmers interviewed in the survey undertaken by Chapman et al.
(2016). Policy 4's key difference from the 3-3-2 system (policy 3) is its
increased exibility, that is, its responsiveness to high ow events,
which maximises sediment deposition and ood protection benets.
Each policy set-up was tested for three different farm sizes (we use
farm size as a proxy for wealth), upper quartile = 3 ha (UQ), median =
1.7 ha, and lower quartile = 1 ha (LQ), based on the survey data collected in An Giang. Each set-up was subjected to Monte-Carlo simulation,
with 100 runs, each inuenced by stochastic, normally distributed, random variation entering the system through the exogenous forcing variables detailed in Section 4.1 and dened by the standard deviation
values derived from secondary data and available in Table 1. This feature
gives the model the power to test the system's dynamic response to
temporal peaks and troughs in exogenous variables.
4.5. Analysis of model outcomes
The SDM results were initially analysed using a set of key indicators
designed to capture differences in dynamics between scenarios; those
indicators included: (i) annual rice production per hectare, (ii) government net prot (rice export revenue assuming an export tax rate of 10%,
based on Pham, 2010, minus any policy costs), (iii) total annual sediment deposition allowed, (iv) farmer mean disposable income (farming
prots minus minimum wage), (v) farmer income stability (mean number of times disposable income dropped to zero per simulation), (vi)
farmer debt (mean total at simulation end), and (vii) prevalence of
debt among farmers (percentage of farmers in greater than USD 50
debt at simulation end). In all cases individual metrics are standardised
(out of 100) using the linear additive model (unweighted) commonly

A. Chapman, S. Darby / Science of the Total Environment 559 (2016) 326338

used for multi-criteria decision analysis (MCDA) and scored on a simplied +/ scale, as is commonly used in system dynamics projects (e.g.
Costa et al. 2011). This scoring system aims to ensure conclusions are
not drawn from the model's outputs which are in fact beyond the
model's capabilities. However, for reference, and particularly to highlight the model's standard errors which are difcult to represent
under this scoring system, all raw results are included in the Supplementary Materials. All results were analysed with a discount rate of
3.5% applied post-hoc, as recommended by the British Government's
(2011) Green Book for time periods spanning 30 years or less, as is
the case here.

5. Results
5.1. Double vs triple-cropping
We rst look at the key changes in socioeconomic system dynamics
resulting from the main transition or adaptation that has already taken
place, i.e. the shift from low to high river dykes associated with the
move from double (policy 2) to triple (policy 1) rice-cropping.
First, and as expected, across all farm sizes Fig. 3 shows that a clear
substitution of sediment (and associated nutrients) for annual rice production (and hence government export prot) takes place when the
double-cropping spin-up period ends (at season 53, marked by the
dashed vertical green line in Fig. 3) and the triple-cropping (policy 1) begins. In Table 7 we compare the system conditions of the farmers (two
thirds of An Giang province) who pursued policy 1 (triple-cropping)
with the farmers who remained double-cropping (policy 2). Table 7 indicates that the shift to triple-cropping has negative outcomes on the
majority of indicators, barring rice production and government prot,
for farms of smallest size (LQ). Results for farms of median size are similar, except that triple-cropping (policy 1) affords greater income stability in this case. However, for farmers with greatest land wealth (i.e. the
UQ of farm sizes), the shift to triple-cropping is predicted to be highly
advantageous, comparatively benetting all indicators bar sediment deposition. The aggregate movements between farm sizes are visualised in
Fig. 4. This comparative analysis highlights the presence of system dynamics which differentiate outcomes between farm sizes; below we explore the internal drivers of this process.
During the model evaluation detailed in Section 4.3.2 we found that
the model systematically overestimated the farming efciency that
triple-cropping farmers were able to achieve. We note that the implications of this inaccuracy for our ndings are simply to strengthen the

Fig. 3. An illustrative run of policy 1 for a farm of median size, without random variation
applied. Shown: the seasonal prot of the farmer (smoothed solid line); the quantity of
sediment-bound nutrients available to crops each season (smoothed dotted line); and
the seasonal rice yield (smoothed dashed line). The three year (nine season) smooth is
applied to improve the visualisation of the double-cropping system which would
otherwise present with one fallow season every three seasons. The chosen indicators
have been converted to Z-Scores, i.e. the number of standard deviations each value is
from the mean.

333

preferentiality of double-cropping over triple-cropping for lower quartile size farms, and to make the net change for median size farms negligible. But, there are no ramications for the comparative conclusions
which can be drawn from the model.
The source of the income and debt penalty imposed by the triplecropping system (policy 1) on poorer (smaller) farmers lies in the increased total and proportional articial fertiliser application required
by the addition of the third crop, and the loss of free sediment-bound
fertilisation. The combination of lost free sediment-bound nutrients
and the addition of a third rice crop results in an average increase in annual articial fertilisation of 87% when comparing double (policy 2) to
triple (policy 1) cropping. The cost of this increased demand for articial
fertilisation is greater on poorer farmers because the model predicts
they operate at a lower level of input efciency. Averaging across the
testing period of our model it is seen that under the triple-cropping system (policy 1) LQ farmers were at approximately a 9% input efciency
disadvantage against UQ farmers whereas, under double-cropping (policy 2), their disadvantage reduces to 5% this difference being a result
of the farmers' increased economic capacity to invest in efciencyenhancing technology under policy 2. The dynamic implication of this
phenomenon is that, when combined with the ner margins that
smaller-scale farmers operate on, triple-cropping (policy 1) farmers
are unable to build up a large enough contingency fund to protect themselves against the model's stochastic fertiliser price spikes. In other
words, the shift to triple-cropping reduces the economic resilience of
poorer farmers in particular. This phenomenon is examined in more detail in Fig. 5. Notably, the debt spikes which result are found almost exclusively later on (beginning around 69 years from policy implementation)
in the simulation meaning that evaluations of the (apparent) success of
the transition to triple-cropping made in the aftermath of the shift
might offer a misleading assessment of its performance over the long
term. Moreover, Fig. 5 hints that once the farmer's economic resilience
is broken, the implications are greater than just a one off loan being
taken out, with debt often recurring in subsequent seasons.
The model also contributes operational insight into the reasons
for the 69 year lag in the onset of debt. In Fig. 3 it can be seen that
for the period between 24 and 69 years after implementation of
triple-cropping farmers enjoy a period of substantial nancial success. This success is harder to detect when the smooth is removed
from the model outputs, as in Fig. 6AB, as it is sourced from the continuous rather than staggered nature of the farmer's prots. Input
costs do not instantly respond to the increased number of rice
crops thanks to the buffer of the nutrient rich deltaic soil that is
maintained by sediment deposition, as shown in the example simulation (Fig. 6B). However, the model suggests that this boost in
prots is inevitably followed by a decline, initially relatively rapidly,
then more slowly, through to the simulation end. This decline can be
linked to the farmer using up the buffer of natural fertilisation provided by sediment deposition and transitioning to what is sometimes termed an open throughput system (Berg, 2002) that is
entirely reliant on articial fertiliser inputs. This, in turn, has signicant economic ramications. As the sediment buffer declines the burden on articial fertilisation increases, and the farmer must then
identify the optimum level of fertilisation in a context of increasing
and variable fertiliser prices (Fig. 6C), a process which lasts beyond
the point at which sediment-bound nutrients are depleted.
Another key factor controlling farmers' success (or otherwise) lies in
their ability to sustain their technological capacity and hence input efciency. The model simulations identify two mechanisms that control the
level of investment in technological advancement. The example simulation in Fig. 6 highlights the rst mechanism: low prot periods (labelled
D) reduce a farmer's ability to invest in their technological capacity and
lead to a slump (labelled E) which, particularly under triple-cropping,
takes some time to recover from. The second factor, which explains
why this slump is harder to recover from under policy 1 and contributed
to the lower technological capacity seen under policy 1 versus policy 2

334

A. Chapman, S. Darby / Science of the Total Environment 559 (2016) 326338

Table 7
Four policies scored comparatively for three farm sizes: LQ: Lower Quartile, M: Median, UQ: Upper Quartile. Scoring is presented on a simplied comparative scale containing ve ratings
(,,O,+,++) where represents the lowest scoring policy and ++ the highest.
Total Rice production

Government Prot

Total sediment deposition

Disposable income

Income stability

Total debt

Debt prevalence

LQ

Triple-cropping
Double-cropping
3-3-2
High year opening

1
2
3
4

++

+
+

++

+
+

++

++

++

++
O

++

Triple-cropping
Double-cropping
3-3-2
High year opening

1
2
3
4

++

+
+

++

+
+

++

++

++

+
+

++

++

UQ

Triple-cropping
Double-cropping
3-3-2
High year opening

1
2
3
4

++

+
+

++

+
+

++

++

++

+
+

++
+

++
O

(for median and LQ size farms), related to the lower prot margins per
rice growing season. Farmers only invest in technology when the seasonal surplus is sufcient for them to feel it worth the risk (the risk
being insufcient backup funds resulting in the future accrual of debt).
Policy 1 involves continuous cropping of low surplus seasons with a
high risk of unexpected input costs; policy 2 involves two high surplus
seasons followed by lower risk of unexpected costs, giving farmers
greater security to invest.
The simulation's suggestion that An Giang farmers are either currently, or likely soon will be, suffering from an increasing debt burden
is one that can be substantiated in the literature and therefore provides
an additional validation of the model. The debt issues faced by poorer
farmers in particular have been mentioned in previous qualitative research (e.g. Garschagen et al., 2012), with one study estimating that
up to 87% of a commune can rely on money lenders (Swain et al.,
2008). Given the relatively recent conversion of many communes to
triple-cropping (while conducting our survey we encountered communes that had converted as recently as 2013), the delay in the onset
of debt may mean this issue has not yet been realised or appreciated
to its fullest extent.
5.2. Implemented variant policies
In this section we use the SDM to explore the effectiveness of a variant rice cultivation strategy that has been implemented to alleviate the
negative impacts described above. Specically, the 33-2 cropping

rotation is the only policy implemented which directly addresses the


role of sediment exclusion in the system, albeit as of 2014 it was only
found in 2 of the 9 An Giang communes surveyed. From Table 7 and
Fig. 4 it can be seen that, comparatively, the 3-3-2 rotation (policy
3) did not excel in any of the key indicators. The 3-3-2 cropping system
attempts to nd a middle ground between the two extremes of high and
low sediment deposition. However, importantly the 3-3-2 rotation involves regular switching between cropping systems, and it is in this regular switching that its poor performance lies. Every third year,
independent of the magnitude of the ood and hence sediment deposition potential experienced in that third year, the farmer sacrices a crop
in order to receive sediment. This means that in the event that the third
year coincides with a low ood, sediment deposition potential is reduced relative to a system that switches responsively to maximise sediment deposition potential. The sediment deposited holds both short
and medium term value. However, the total short term value the farmer
gains from the sediment, combined with income from off-season activities, is lower than the income they would gain from a season of rice (although, when the medium term gain is considered the net gain may be
greater). The triple-cropping system requires that the farmer hold sufcient (and signicant) short-term funds to cover the increased
fertilisation costs. Examining the model simulations we found that the
lower income obtained in the short-term during the double-cropping
season reduces the contingency funds available to the farmer for the
next year's fertilisation costs. Thus, farmers struggle to fund their
input costs in the cropping seasons directly following a fallow season,
and their resilience to high fertiliser prices therefore declines. Fig. 7,
which shows the probabilities of debt being incurred at different stages,
illustrates this issue. As a result, the 3-3-2 system (policy 3) ranks poorly
in terms of debt prevalence. Again, this phenomenon only tends to have
an impact after the initial income boost the farmer receives from
switching to triple-cropping subsides, which our model indicates occurs
after 69 years. However, it should be noted that the precise length of
this lag is primarily inuenced by the parameter which determines
the time between nutrient deposition and its availability for plant
uptake; this parameter was one of ve identied as having a notably
weaker evidence base in Section 4.3.1. In contrast to policy 1, where
triple-cropping performs notably worse for land-poor farmers, the 3-32 system's performance is uniformly poor across all farm size categories.
5.3. Proposed variant policies

Fig. 4. The comparative score (out of 100) of each policy in each indicator has been
aggregated for each of the policies and for the three farm size classes. Comparisons can
be made between the performances of each policy at different farm sizes (lower quartile
(LQ), median (M) and upper quartile (UQ) of the 195 farms surveyed by Chapman et al.
(2016). Indicators have not been weighted.

Here we ask: how do other, proposed, policies affect the system's


dynamics and how do they compare with the status quo? To address
this question we tested a model set-up (policy 4) in which the farmer
performed double-cropping and allowed inundation and sediment
deposition only in years with higher sediment deposition potential.
Theoretically this policy would ensure the fallow season was optimised
for maximum benet and we hypothesised that this would reduce some

A. Chapman, S. Darby / Science of the Total Environment 559 (2016) 326338

335

Fig. 5. Four examples of peaks in fertiliser prices. A, B, and C show LQ simulations, D shows a UQ simulation. In A and B we see peaks in fertiliser prices causing debt spikes. Commonly these
spikes occur later in the simulation and the initial spike tends to have a knock-on effect on subsequent seasons. Graph C is a rare example of a small early-simulation debt spike caused by
two localised price spikes that does not have a knock-on effect. In graph D we see an example of a UQ size farmer coping with a severe price spike without incurring debt.

of the negative traits of the 3-3-2 rotation (policy 3). Indeed, when all
indicators are aggregated policy 4 is seen to offer an improvement relative to policy 3 (Fig. 4). However, policy 4 results in farmers incurring
marginally greater debt problems (Table 7). The phenomenon causing
this is now familiar. While fallow seasons are smaller in number

under policy 4 versus policy 3, the random nature of when peak ood
events occur occasionally means successive years of double-cropping.
Such occurrences have a signicant impact on farmers' economic reserves and result in debt in the subsequent season in which the farmer
returned to triple-cropping.

Fig. 6. Four simulation indicators (seasonal prot, fertiliser applied, yield, and input efciency) shown for a sample run of a farm of median size operating Policy 1 with random variation on
and no smoothing applied. In green, the point at which the switch in cropping pattern took place is highlighted. The chosen indicators have been converted to Z-Scores, i.e. the number of
standard deviations each value is from the mean. Some key features are labelled.

336

A. Chapman, S. Darby / Science of the Total Environment 559 (2016) 326338

Fig. 7. The probability of a median size farmer falling into debt during the policy simulation
period on a given season under policy 3. For comparison the average probabilities for the
triple and double cropping policies are shown.

Despite the above caveat, when all of our indicators are aggregated,
policy 4 performs better than policy 3 across all farm size categories
(Fig. 4). This is a direct consequence of the optimisation of sedimentbound nutrient potential, which improves the long term yield to
fertiliser ratio. Furthermore, while opinions differ on the best way to
protect local livelihoods from intensifying uvial oods, most agree
that controlled inundation during intense ood events is an effective
mechanism, particularly for protecting livelihoods downstream of the
paddy compartments. This additional benet would further increase
the preferentiality of policy 4 against the alternatives if it were included
in our aggregate scores (Fig. 4).

offers a short-term boost in income, farmers face a signicant risk of a


decline in productivity and protability subsequently. Indeed, our evidence suggests that from a local, medium to long-term, perspective
there is negligible benet for the majority of farmers making the
triple-cropping shift. The benets of triple-cropping are instead felt
overwhelmingly at a macro-economic scale through the benets for
total rice production, the governmental export income this generates,
and in the nancial gains made by the land-wealthier farmers with
the margins and contingency to cope with shocks in fertiliser prices.
Policy makers therefore need to weigh these benets against the ood
protection and delta sustainability advantages of alternative policies
(such as policy 4, the strategic ooding of paddy compartments during
high ood years) which facilitate sediment deposition and offer better
outcomes for poorer farmers. In these respects our study provides valuable detail and support to the ndings of the Mekong Delta Plan (2013)
and specically its recommendations to implement strategic controlled
ooding of northern regions of the Vietnamese Mekong delta (our policy 4, Fig. 4). Moreover, it builds on a body of literature (e.g. Szabo et al.,
2015) emphasising the important role of wealth inequality in determining society's ability to cope with environmental change, and importantly also adaptations, in delta regions.
The Mekong delta (and our case study province, An Giang) is a key
region in terms of its contribution to global food security, its large
population, and the intensity of the environmental change and development pressures it is subject to. Achieving wellbeing for all those dependent on such regions is a major policy challenge (Dearing et al., 2014).
Our exploration of the dynamics of different policies and their implications across multiple objectives for the Mekong delta system highlights
the difculty of balancing different objectives to nd what Dearing et al.
(2014) term safe and just operating spaces for society. Nevertheless,
our use of a novel system dynamics approach towards adaptation decision making, in a context of double exposure to climate change and development impacts in river deltas, offers a step forward towards
meeting this challenging objective.

6. Discussion and conclusion


In this paper we have established, tested and applied a new model
capable of analysing different adaptation policies in terms of how their
different dynamics encourage or discourage a just and resilient system
in deltas. Particularly, to our knowledge, this is the rst such model to
integrate the socioeconomic role of uvial sediment deposition. The
new model performs well in simulating observed behaviours of the system, substantiating and providing quantitative, operational, evidence
towards concerns about the prevalence of debt in the adapted ricecropping system of An Giang. With this study we have therefore
shown the operability of a system dynamics approach which evaluates
disparate cross-disciplinary factors controlling adaptation success. We
argue that this approach has wide transferability, offering the potential
for rapid systems assessment in regions facing similarly intense changes, and particularly the delta context, which is of high importance to
global food security. Notable candidate systems for the application of
this model might be found in the other South/Southeast Asian deltas
under threat, such as the Irrawaddy, Chao Phraya, and GangesBrahmaputra (Auerbach et al., 2015).
A key outcome of our modelling is to highlight an operational mechanism through which the loss of uvial sediment, combined with lower
input efciency and high and variable fertiliser prices, can force a greater debt burden on poorer farmers operating the triple-cropping rice
cultivation system. The system has been the focus of rapid recent expansion, and hence this nding raises signicant concern of a reduction in
the resilience of poorer groups to future fertiliser price growth and
spikes. Furthermore, we present evidence that through these mechanisms the high dyke adaptation which underpins the triple-cropping
system may actively increase the efciency gap, and hence wealth
gap, between the wealthiest farmers and the rest of agricultural society.
Our ndings suggest that although the recent switch to triple-cropping

Acknowledgements
The authors wish to extend their thanks to Professor P.I. Davidsen
and the System Dynamics group at the University of Bergen for their
support with the modelling process, and the University of Southampton
for PhD funding. S.E.D's contribution to this paper was supported by
award NE/JO21970/1 from the UK Natural Environment Research Council (NERC).
Appendix A. Supplementary data
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.scitotenv.2016.02.162.
References
AGSO, 2013. Statistical Yearbook of An Giang: 2012. An Giang Statistical Ofce, Long
Xuyen.
Alcal, F.J., Martnez-Valderrama, J., Robles-Marn, P., Guerrera, F., Martn-Martn, M.,
Raffaelli, G., ... Asebriy, L., 2015. A hydrological-economic model for sustainable
groundwater use in sparse-data drylands: application to the Amtoudi Oasis in southern Morocco, northern Sahara. Sci. Total Environ. 537, 309322. http://dx.doi.org/10.
1016/j.scitotenv.2015.07.062.
Anthony, E.J., Brunier, G., Besset, M., Goichot, M., Dussouillez, P., Nguyen, V.L., 2015.
Linking rapid erosion of the Mekong River delta to human activities. Sci. Rep. 5,
14745. http://dx.doi.org/10.1038/srep14745.
Auerbach, L.W., Goodbred Jr., S.L., Mondal, D.R., Wilson, C.a., Ahmed, K.R., Roy, K., Steckler,
M.S., Small, C., Gilligan, J.M., Ackerly, B.a., 2015. Flood risk of natural and embanked
landscapes on the GangesBrahmaputra tidal delta plain. Nat. Clim. Chang. 5,
153157.
Barnett, J., O'Neill, S., 2010. Maladaptation. Glob. Environ. Chang. 20 (2), 211213.
Berg, H., 2002. Rice monoculture and integrated rice-sh farming in the Mekong Delta,
Vietnameconomic and ecological considerations. Ecol. Econ. 41 (1), 95107.
http://dx.doi.org/10.1016/S0921-8009(02)00027-7.

A. Chapman, S. Darby / Science of the Total Environment 559 (2016) 326338


Birkmann, J., Garschagen, M., Tuan, V. Van, 2012. Vulnerability, coping and adaptation to
water related hazards in the Vietnamese Mekong Delta. In: Renaud, F.G., Kuenzer, C.
(Eds.), The Mekong Delta System. Springer Netherlands, Dordrecht. http://dx.doi.org/
10.1007/978-94-007-3962-8.
Bosma, R.H., Udo, H.M.J., Verreth, J.A.J., Visser, L.E., Nam, C.Q., 2005. Agriculture diversication in the Mekong Delta: farmers' motives and contributions to livelihoods.
Asian J. Agric. Dev. 2 (1&2), 4966.
British Government, 2011. The Green Book: Appraisal and Evaluation in Central Government. H.M.Treasury, London. http://greenbook.treasury.gov.uk/index.htm.
Chapman, A., Darby, S., Hong, H., Tompkins, E., Van, P.D.T., 2016. Clim. Chang. http://dx.
doi.org/10.1007/s10584-016-1684-3.
Coleman, J., Huh, O., 2003. Major Deltas of the World: A Perspective from Space. Baton
Rouge, LA, USA.
Costa, M.A.M., Moors, E.J., Fraser, E.D.G., 2011. Socioeconomics, policy, or climate change:
what is driving vulnerability in Southern Portugal? Ecol. Soc. 16 (1).
Dace, E., Muizniece, I., Blumberga, A., Kaczala, F., 2015. Searching for solutions to mitigate
greenhouse gas emissions by agricultural policy decisions application of system dynamics modeling for the case of Latvia. Sci. Total Environ. 527-528, 8090. http://dx.
doi.org/10.1016/j.scitotenv.2015.04.088.
Day, J.W., Barras, J., Clairain, E., Johnston, J., Justic, D., Kemp, G.P., ... Yaez-Arancibia, A.,
2005. Implications of global climatic change and energy cost and availability for the
restoration of the Mississippi delta. Ecol. Eng. 24 (4 SPEC. ISS.), 253265. http://dx.
doi.org/10.1016/j.ecoleng.2004.11.015.
Dearing, J.A., Wang, R., Zhang, K., Dyke, J.G., Haberl, H., Hossain, M.S., ... Poppy, G.M., 2014.
Safe and just operating spaces for regional social-ecological systems. Glob. Environ.
Chang. 28, 227238. http://dx.doi.org/10.1016/j.gloenvcha.2014.06.012.
Demombynes, G., Linh, H.V., 2015. Demystifying Poverty Measurement in Vietnam.
Hanoi, Vietnam.
Evers, H., Benedikter, S., 2009. Hydraulic bureaucracy in a modern hydraulic societystrategic group formation in the Mekong Delta, Vietnam. Water Altern. 2 (3),
416439.
FAO, 2014. An Giang Rice Price Data. Food and Agriculture Organization of the United Nations (Retrieved November 17, 2014, from http://www.fao.org/giews/pricetool/).
Ford, A., 2010. Modeling the Environment. Island Press, Washington, DC.
Fraser, E.D.G., Dougill, A.J., Hubacek, K., Quinn, C.H., Sendzimir, J., Termansen, M., 2011.
Assessing vulnerability to climate change in dryland livelihood systems: conceptual
challenges and interdisciplinary solutions. Ecol. Soc. 16 (3). http://dx.doi.org/10.
5751/ES-03402-160303.
Fssel, H.-M., 2007. Adaptation planning for climate change: concepts, assessment approaches, and key lessons. Sustain. Sci. 2 (2), 265275. http://dx.doi.org/10.1007/
s11625-007-0032-y.
Garschagen, M., Revilla Diez, J., Dang, N., Kraas, F., 2012. Socio-economic development in
the Mekong Delta: between the prospects for progress and the realms of reality.
In: Renaud, F.G., Kuenzer, C. (Eds.), The Mekong Delta System. Springer
Netherlands, Dordrecht, pp. 133165. http://dx.doi.org/10.1007/978-94-0073962-8.
Gies, L., Agusdinata, D.B., Merwade, V., 2014. Drought adaptation policy development and
assessment in East Africa using hydrologic and system dynamics modeling. Nat. Hazards 74 (2), 789813. http://dx.doi.org/10.1007/s11069-014-1216-2.
GSO, 2014. General Statistics Ofce of Vietnam: statistical data. (Retrieved January 15,
2014, from) http://www.gso.gov.vn/default_en.aspx?tabid=491.
Hensel, P.F., Day, J.W., Pont, D., Day, J.N., 1998. Short-term sedimentation dynamics in the
Rhone River Delta, France: the importance of riverine pulsing. Estuaries 21 (1), 52.
http://dx.doi.org/10.2307/1352546.
Hood, G., 2004. Indirect environmental effects of dikes on estuarine tidal channels: thinking outside of the dike for habitat restoration and monitoring. Estuaries 27 (2),
273282. http://dx.doi.org/10.1007/BF02803384.
Hueglas, Z., Templeton, D., 2010. Addition of crop management technology and costefciency impacts: the case of three reductions, three gains in the Mekong River
Delta of Vietnam. In: Palis, F.G., Singleton, G.R., Casimero, M.C., Hardy, B. (Eds.), Research to Impact: Case Studies for Natural Resource Manage for Irrigated Rice in
Asia. International Rice Research Institute (IRRI), Los Banos (Phillipines), p. 370.
Hung, N.N., Delgado, J.M., Gntner, A., Merz, B., Brdossy, A., Apel, H., 2014. Sedimentation
in the oodplains of the Mekong Delta, Vietnam part II: deposition and erosion.
Hydrol. Process. 28 (7), 31453160. http://dx.doi.org/10.1002/hyp.9855.
Ibez, C., Canicio, A., Day, J.W., Curco, A., 1997. Morphologic development, relative sea
level rise and sustainable management of water and sediment in the Ebro Delta,
Spain. J. Coast. Conserv. 3 (2), 191202 (Retrieved from http://www.jstor.org/
stable/25098268).
Ibez, C., Day, J.W., Reyes, E., 2014. The response of deltas to sea-level rise: natural mechanisms and management options to adapt to high-end scenarios. Ecol. Eng. http://dx.
doi.org/10.1016/j.ecoleng.2013.08.002.
ISG-MARD, 2011. Executive summary. Vietnam Rice, farmers and Rural Development: From Successful Growth to Sustainable Prosperity. International Support
Group (ISG) of the Ministry of Agriculture and Rural Development (MARD),
Vietnam, Can Tho, Vietnam (Retrieved from http://www.isgmard.org.vn/VHForums/
VHViewTopic.asp?id=4#_null).
Khai, H., Yabe, M., 2011. Technical efciency analysis of rice production in Vietnam.
J. ISSAAS 17 (1), 135146 (Retrieved from http://issaas.org/journal/v17/01/journalissaas-v17n1-15-khai_yabe.pdf).
Kondolf, G.M., Rubin, Z.K., Minear, J.T., 2014. Dams on the Mekong: cumulative sediment
starvation. Water Resour. Res. 50, 51585169. http://dx.doi.org/10.1002/
2013WR014651.Received.
Kontgis, C., Schneider, A., Ozdogan, M., 2015. Remote sensing of environment mapping
rice paddy extent and intensication in the Vietnamese Mekong River Delta with

337

dense time stacks of Landsat data. Remote Sens. Environ. 169, 255269. http://dx.
doi.org/10.1016/j.rse.2015.08.004.
Kuenzer, C., Renaud, F.G., 2012. Climate and environmental change in river deltas globally: expected impacts, resilience, and adaptation. In: Renaud, F.G., Kuenzer, C. (Eds.),
The Mekong Delta System. Springer Environmental Science and Engineering, Dordrecht, pp. 133165. http://dx.doi.org/10.1007/978-94-007-3962-8.
Kummu, M., Lu, X.X., Wang, J.J., Varis, O., 2010. Basin-wide sediment trapping efciency of
emerging reservoirs along the Mekong. Geomorphology 119 (34), 181197. http://
dx.doi.org/10.1016/j.geomorph.2010.03.018.
Laborte, A.G., de Bie, K.C.A.J.M., Smaling, E.M.A., Moya, P.F., Boling, A.A., Van Ittersum, M.K.,
2012. Rice yields and yield gaps in Southeast Asia: past trends and future outlook.
Eur. J. Agron. 36 (1), 920. http://dx.doi.org/10.1016/j.eja.2011.08.005.
Lempert, R., Nakicenovic, N., Sarewitz, D., Schlesinger, M., 2004. Characterizing climatechange uncertainties for decision-makers an editorial essay. Clim. Chang. 65
(12), 19. http://dx.doi.org/10.1023/B:CLIM.0000037561.75281.b3.
Lempert, R.J., Groves, D.G., Popper, S.W., Bankes, S.C., 2006. A general, analytic method for
generating robust strategies and narrative scenarios. Manag. Sci. 52 (4), 514528.
http://dx.doi.org/10.1287/mnsc.1050.0472.
Liang, X.Q., Chen, Y.X., Nie, Z.Y., Ye, Y.S., Liu, J., Tian, G.M., ... Tuong, T.P., 2013. Mitigation of
nutrient losses via surface runoff from rice cropping systems with alternate wetting
and drying irrigation and site-specic nutrient management practices. Environ. Sci.
Pollut. Res. 20 (10), 69806991.
Manh, N.V., Dung, N.V., Hung, N.N., Merz, B., Apel, H., 2014. Large-scale suspended sediment transport and sediment deposition in the Mekong Delta. Hydrol. Earth Syst.
Sci. 18 (8), 30333053. http://dx.doi.org/10.5194/hess-18-3033-2014.
MARD, 2008. Issuance of the Action Plan Framework for Adaptation to Climate Change in
the Agriculture and Rural Development Sector Period 2008-2020. Hanoi.
MDP, 2013. Mekong Delta Plan: Long-term Vision and Strategy for a Prosperous and Sustainable Delta.
Mohanty, S., Wassmann, R., Nelson, A., Moya, P., Jagadish, S., 2013. Rice and Climate Change:
Signicance for Food Security and Vulnerability (No. 49). Los Baos (Philippines).
MPI, 2006. The Five-year Socio-economic Development Plan 20062010. Hanoi.
Nixon, S.W., 2003. Replacing the Nile: are anthropogenic nutrients providing the fertility
once brought to the Mediterranean by a great river? Ambio 32 (1), 3039. http://dx.
doi.org/10.1639/0044-7447(2003)032{[}0030:RTNAAN]2.0.CO;2.
O'Brien, K.L., Leichenko, R.M., 2000. Double exposure: assessing the impacts of climate
change within the context of economic globalization. Glob. Environ. Chang. 10 (3),
221232. http://dx.doi.org/10.1016/S0959-3780(00)00021-2.
Olde Venterink, H., Vermaat, J.J.E., Pronk, M., Wiegman, F., van der Lee, G.E.M., van den
Hoorn, M.W., ... Verhoeven, J.T.A., 2006. Importance of sediment deposition and denitrication for nutrient retention in oodplain wetlands. Appl. Veg. Sci. 9 (2),
163174. http://dx.doi.org/10.1111/j.1654-109X.2006.tb00665.x.
Oppenheimer, M., Campos, M., Warren, R., Birkmann, J., Luber, G., O'Neill, B.C., Takahashi,
K., 2014. Emergent risks and key vulnerabilities. In: Field, C.B., Barros, V.R., Dokken,
D.J., Mach, K.J., Mastrandrea, M.D., Bilir, T.E., White, L.L. (Eds.), Climate Change
2014: Impacts, Adaptation, and Vulnerability. Part A: Global and Sectoral
AspectsContribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel of Climate Change. Cambridge University Press, Cambridge,
United Kingdom and New York, NY, USA, pp. 10391099.
Pham, H.N., 2010. The Vietnamese rice industry during the global food crisis. In: Dawe, D.
(Ed.), The Rice Crisis: Markets, Policies and Food Security. FAO and Earthscan,
London, p. 393. http://dx.doi.org/10.4324/9781849776684.
Pham, C.H., 2011. Dikes and Farmers: Politics, Economics and Environmental Impacts of
Dyke Construction in the Mekong Delta/Vietnam. Lit Verlag, Zurich (Retrieved from
http://books.google.co.uk/books?id=iL3IaAgovLIC&pg=PA176&lpg=PA176&dq=
Floods+and+Farmers:+Politics,+Economics,+and+Environmental+Impacts+
of+Dyke+Construction+in+the+Mekong+Delta/Vietnam&source=bl&ots=
pPOpTa_pOG&sig=WeIEcjlW4vvPh-av2zDtFVMNd-Y&hl=en&sa=X&e).
Pham, T., Tran, T., Tran, H., Trinh, K., Nguyen, H., Le, D., ... Dobermann, A., 2004. Sitespecic nutrient management in irrigated rice systems of the Mekong Delta of
Vietnam. In: Dobermann, A., Witt, C., Dawe, D. (Eds.), Increasing Productivity
of Intensive Rice Systems Through Site-specic Nutrient Management, p. 410
(Eneld, N.H. (USA)).
Phong, L.T., Stoorvogel, J.J., van Mensvoort, M.E.F., Udo, H.M.J., 2011. Modeling the soil nutrient balance of integrated agriculture-aquaculture systems in the Mekong Delta, Vietnam.
Nutr. Cycl. Agroecosyst. 90 (1), 3349. http://dx.doi.org/10.1007/s10705-010-9410-4.
Ray, D.K., Ramankutty, N., Mueller, N.D., West, P.C., Foley, J.a., 2012. Recent patterns of
crop yield growth and stagnation. Nat. Commun. 3, 1293. http://dx.doi.org/10.1038/
ncomms2296.
Reardon, T., Chen, K.Z., Minten, B., Adriano, L., Dao, T.A., Wang, J., Gupta, S. Das, 2014. The
quiet revolution in Asia's rice value chains. Ann. N. Y. Acad. Sci. 1331 (1), 106118.
http://dx.doi.org/10.1111/nyas.12391.
Rovira, A., Ibez, C., 2007. Sediment management options for the lower ebro river and its
delta. J. Soils Sediments 7 (5), 285295. http://dx.doi.org/10.1065/jss2007.08.244.
Rutten, M., van Dijk, M., van Rooij, W., Hilderink, H., 2014. Land use dynamics, climate
change, and food security in Vietnam: a global-to-local modeling approach. World
Dev. 59, 2946. http://dx.doi.org/10.1016/j.worlddev.2014.01.020.
Sakamoto, T., Phung, V.C., Nhan, V.N., Kotera, A., Yokozawa, M., 2009. Agro-ecological interpretation of rice cropping systems in ood-prone areas using MODIS imagery.
Photogramm. Eng. Remote. Sens. 75 (4), 413424.
Seto, K.C., Woodcock, C.E., Song, C., Huang, X., Lu, J., Kaufmann, R.K., 2002. Monitoring
land-use change in the Pearl River Delta using Landsat TM. Int. J. Remote Sens. 23
(10), 19852004. http://dx.doi.org/10.1080/01431160110075532.
Shrestha, B., Babel, M.S., Maskey, S., van Griensven, A., Uhlenbrook, S., Green, A.,
Akkharath, I., 2013. Impact of climate change on sediment yield in the Mekong

338

A. Chapman, S. Darby / Science of the Total Environment 559 (2016) 326338

River basin: a case study of the Nam Ou basin, Lao PDR. Hydrol. Earth Syst. Sci. 17 (1),
120. http://dx.doi.org/10.5194/hess-17-1-2013.
Simonovic, S.P., Li, L.H., 2004. Sensitivity of the Red River basin ood protection system to
climate variability and change. Water Resour. Manag. 18 (2), 89110. http://dx.doi.
org/10.1023/B:WARM.0000024702.40031.b2.
SIWRR, 2013. Suspended Sediment Concentration Data Provided by the Southern Institute of Water Resources Research, Ho Chi Minh City.
Smajgl, A., Toan, T.Q., Nhan, D.K., Ward, J., Trung, N.H., Tri, L.Q., ... Vu, P.T., 2015.
Responding to rising sea levels in the Mekong Delta. Nat. Clim. Chang. 5, 167174.
http://dx.doi.org/10.1038/NCLIMATE2469.
Sterman, J.D., 2000. Business Dynamics: Systems Thinking and Modeling for a Complex
World. Irwin/McGraw-Hill, Boston.
Suckall, N., Tompkins, E., Stringer, L., 2014. Identifying trade-offs between adaptation,
mitigation and development in community responses to climate and socioeconomic stresses: evidence from Zanzibar, Tanzania. Appl. Geogr. 46, 111121.
http://dx.doi.org/10.1016/j.apgeog.2013.11.005.
Sunik, J., Vamvakeridou-Lyroudia, L.S., Savi, D.A., Kapelan, Z., 2013. Integrated modelling
of a coupled water-agricultural system using system dynamics. J. Water Clim. Chang.
4 (3), 209231.
Swain, R.B., Van Tuan, V., Van Sanh, N., 2008. Micronance and poverty reduction in the
Mekong Delta in Vietnam. Afr. Asian Stud. 7 (2), 191215. http://dx.doi.org/10.
1163/156921008X318736.
Syvitski, J.P.M., Kettner, A., 2011. Sediment ux and the Anthropocene. Philos. Trans. A
Math. Phys. Eng. Sci. 369 (1938), 957975. http://dx.doi.org/10.1098/rsta.2010.
0329.
Syvitski, J.P.M., Kettner, A.J., Overeem, I., Hutton, E.W.H., Hannon, M.T., Brakenridge, G.R.,
... Nicholls, R.J., 2009. Sinking deltas due to human activities. Nat. Geosci. 2 (10),
681686. http://dx.doi.org/10.1038/ngeo629.
Szabo, S., Hossain, M.S., Adger, W.N., Matthews, Z., Ahmed, S., Lzr, A.N., Ahmad, S., 2015. Soil
salinity, household wealth and food insecurity in tropical deltas: evidence from southwest coast of Bangladesh. Sustain. Sci. http://dx.doi.org/10.1007/s11625-015-0337-1.
Temmerman, S., Kirwan, M.L., 2015. Building land with a rising sea. Science 349 (6248),
588589.
Tessler, Z.D., Vrsmarty, C.J., Grossberg, M., Gladkova, I., Aizenman, H., Syvitski, J.P.M.,
Foufoula-Georgiou, E., 2015. Proling risk and sustainability in coastal deltas of the
world. Science 349 (6248), 638643.
Tin, H.Q., Struik, P.C., Price, L.L., Be, T.T., 2008. Comparative analysis of local and improved
practices used by farmer seed production schools in Vietnam. Field Crop Res. 108,
212221. http://dx.doi.org/10.1016/j.fcr.2008.05.005.

Tran, T.T., 2014. Fertilizer in Vietnam. International Food Policy Research Institute, Hanoi.
Tran, T.U., Kajisa, K., 2006. The impact of green revolution on rice production in Vietnam.
Dev. Econ. 44 (2), 167189. http://dx.doi.org/10.1111/j.1746-1049.2006.00012.x.
Tran, T.U., Hossain, M., Janaiah, A., 2000. Modem farm technology in Vietnam and infrastructure impact. Econ. Polit. Wkly. 35 (52), 46384643.
Tsheboeng, G., Bonyongo, M., Murray-Hudson, M., 2014. Flood variation and soil nutrient
content in oodplain vegetation communities in the Okavango Delta. S. Afr. J. Sci. 110
(3/4), 15. http://dx.doi.org/10.1590/sajs.2014/20130168.
Van, P.D.T., Popescu, I., van Griensven, A., Solomatine, D.P., Trung, N.H., Green, A., 2012. A
study of the climate change impacts on uvial ood propagation in the Vietnamese
Mekong Delta. Hydrol. Earth Syst. Sci. 16 (12), 46374649. http://dx.doi.org/10.
5194/hess-16-4637-2012.
Vietnamese Government, 2011. National strategy on climate change. (Retrieved January
30, 2014, from) http://www.chinhphu.vn/portal/page/portal/English/strategies/
strategiesdetails?categoryId=30&articleId=10051283.
Vietnamese Government, 2013. Decree No. 182/2013/ND-CP: 2014 Minimum Wage.
Hanoi.
Warner, K., 2010. Global environmental change and migration: governance challenges.
Glob. Environ. Chang. Hum. Policy Dimens. 20 (3, SI), 402413. http://dx.doi.org/10.
1016/j.gloenvcha.2009.12.001.
Wassmann, R., Jagadish, S.V.K., Sumeth, K., Pathak, H., Howell, G., Ismail, A., ... Heuer, S.,
2009. Regional vulnerability of climate change impacts on Asian rice production and
scope for adaptation. Adv. Agron. 102, 91133. http://dx.doi.org/10.1016/S00652113(09)01003-7.
Witt, C., Dobermann, A., Abdulrachman, S., Gines, H.C., Guanghuo, W., Nagarajan, R., ... Olk,
D.C., 1999. Internal nutrient efciencies of irrigated lowland rice in tropical and subtropical Asia. Field Crop Res. 63 (2), 113138. http://dx.doi.org/10.1016/S03784290(99)00031-3.
World Bank, 2014. Overview of Commodity Markets. World Bank (Retrieved November 17, 2014, from http://econ.worldbank.org/WBSITE/EXTERNAL/EXTDEC/
EXTDECPROSPECTS/0,contentMDK:21574907~menuPK:7859231~pagePK:641
65401~piPK:64165026~theSitePK:476883,00.html).
Zhang, X.H., Zhang, H.W., Chen, B., Chen, G.Q., Zhao, X.H., 2008. Water resources planning
based on complex system dynamics: a case study of Tianjin city. Commun. Nonlinear
Sci. Numer. Simul. 13 (10), 23282336. http://dx.doi.org/10.1016/j.cnsns.2007.05.
031.